Source-linked AI summary
Feeling Wistful: Reflecting on Scholarly Sensibilities with Creative Reading Traces
Sophia W. Liu, Kate Chier, Shm Garanganao Almeda, Max Kreminski, Bjoern Hartmann
TL;DR
Researchers often explore literature before they can articulate what they seek, while AI increasingly mediates early scholarly encounters. This paper examines curiosity-driven reading and introduces Wistful, a probe that preserves reading paths as creative reading traces. Researchers pursued different paths through a shared landscape, and traces made those differences available for reflection and comparison.
Problem
Researchers often read before they can articulate what they seek, and AI increasingly mediates early encounters with literature, making it important to understand how scholarly sensibilities are expressed through reading.
Source: §1 · p. 2Method
Results
Researchers approached the shared scholarly landscape through different personal reference frames, with experience and interests shaping what they pursued, and traces made these differences available for reflection and comparison.
Source: §1 · p. 2, §7 · p. 12Takeaways & Limitations
Creative reading traces can preserve fleeting expressions of scholarly sensibilities for reflection and exchange, and provide material for studying how those sensibilities appear through reading.
Source: §1 · p. 2, §8 · p. 15The study captures brief expressions in a fixed UIST corpus and 15-minute tasks; it cannot make causal claims about career stage or sensibility development.
Source: §7.5 · p. 14
Abstract
from arXiv · showhide
Researchers often read before they can articulate what they are looking for. As AI increasingly mediates scholarly search and synthesis, understanding and preserving the idiosyncratic judgments guiding early exploration become important. We call these evolving orientations scholarly sensibilities. To understand curiosity-driven reading, we first examined Wikipedia rabbitholing, a self-directed browsing practice, then designed Wistful, a research probe for open-ended scholarly exploration that captures reading paths as creative reading traces. We studied Wistful with 16 HCI researchers---eight junior and eight senior---in comparison with their usual workflows. Researchers approached the same scholarly landscape differently, with familiarity and personal interests shaping what they pursued and semantic proximity and scholarly links shaping their paths. Their traces made these differences visible and let readers revisit and compare their paths. We position creative reading traces as artifacts for reflection and exchange and as a means of studying how scholarly sensibilities are expressed through reading.
1 Introduction
The paper studies how researchers’ emerging scholarly sensibilities shape open-ended reading, especially before they can articulate a precise query. It introduces Wistful, which preserves reading paths as creative reading traces, and examines how those traces make researchers’ different orientations visible for reflection and exchange.
1 Introduction
- Scholarly sensibilities are evolving orientations that guide what researchers notice and judge worth pursuing, expressed in their idiosyncratic reading paths.Researchers draw on prior knowledge and personal experience as a direction takes shape.Source: §1 · p. 2
- Early scholarly reading often precedes a clear query, while AI mediation may shape what researchers encounter and pursue during this formative stage [1] [4].The paper frames understanding how sensibilities appear and develop through reading as a way to study AI-mediated tools over time.Source: §1 · p. 2
- Creative reading traces preserve partial representations of how readers notice, follow, and connect literature, making reading paths available for revisitation and interpretation [53] [55].The paper presents these traces as material for studying scholarly sensibilities.Source: §1 · p. 2
- Wistful brings curiosity-led exploration into scholarly reading, preserving paths and annotations for revisiting and sharing.The study with 16 HCI researchers found that experience and personal interests shaped what they pursued, while traces supported reflection and comparison of paths.Source: §1 · p. 2
2 Related Work
The related work frames scholarly sensibilities as orientations developed through expertise and reading, and scholarly exploration as movement shaped by both readers and their information terrain. It motivates creative reading traces as records that make exploratory processes available for reflection and study.
2.1 Developing Scholarly Ways of Seeing
- Scholarly sensibilities describe evolving orientations toward what researchers notice and judge worth pursuing, shaped by experience and social context.Professional vision and scientific taste provide related accounts of how expertise guides perception and judgment.Source: §2.1 · pp. 2–3
- Experience can shape how flexibly researchers reformulate queries and judge relevance, while emerging research needs may precede the vocabulary to express them.Sensibilities may guide what feels worth pursuing before those judgments can be phrased as a prompt or query.Source: §2.1 · p. 3
2.2 Scholarly Exploration and Mediated Attention
- Scholarly exploration is shaped by evolving interests and the terrain’s available relationships; encounters may in turn reshape the reader’s sensibilities [53].Generative AI may shape attention and directions pursued, while concerns about similar ideas and sources motivate avoiding model-favored directions [1] [4].Source: §2.2 · p. 3
2.3 Trails and Traces for Reflection
- Creative reading traces treat partial records of exploratory choices as meaningful artifacts for reflection, recollection, and study [36, 53, 55].Shared histories can expose others’ paths, supporting reflection and socially cultivated scholarly ways of seeing.Source: §2.3 · pp. 3–4
3 Formative Study
The formative study examined self-directed Wikipedia browsing to understand curiosity-led exploration and inform scholarly reading support. It found that questions emerged along the path, while prior knowledge, personal interests, and remembered moments shaped what readers pursued and how they reconstructed their browsing.
3 Formative Study
3.1 Procedure
- Nine participants browsed for 15 minutes, freely following links, searching, and backtracking, then reconstructed their explorations from memory.Starting articles were selected in relation to participants’ stated interests; interviews also covered prior knowledge and browsing practices.Source: §3 · p. 4, §3.1 · p. 4
- Three authors used open coding on navigation traces, think-aloud data, and interviews, iteratively refining themes through discussion.Two authors coded each participant, and emerging themes guided theoretical sampling.Source: §3.1 · p. 4
3.2 Key Formative Study Findings
- Participants visited 2–21 unique articles in 15 minutes (median 15), with paths ranging from linear sequences to branching and looping explorations.The study traced what shaped pursued directions and what remained meaningful afterward.Source: §3.2 · p. 4
3.2.1 Inquiry Emerged Through the Path
- Questions emerged from encountered words, vague memories, and old questions, with temporary branches sometimes returning to an earlier thread.P6 followed a connection from Monte Carlo methods to Manhattan Project history before returning to the original topic.Source: §3.2.1 · p. 4
3.2.2 Curiosity Reflected Personal Ways of Seeing
- Familiarity, personal interests, backgrounds, and imagined future identities shaped what participants noticed and pursued.Some followed familiar topics or personally resonant connections, while another deliberately sought less familiar territory.Source: §3.2.2 · p. 4
3.2.3 Readers Reconstructed Both Paths and Moments
- Readers recalled both broad path structures and particular moments, motivating multiple views of navigation history for reflection.Participants imagined trees, graphs, nested or overlapping circles, and spatial clusters when representing their browses.Source: §3.2.3 · p. 4
4 Wistful
Informed by a formative study of curiosity-driven browsing, Wistful is a probe for open-ended scholarly exploration that captures paths for revisiting and sharing. Its common scholarly landscape and trace tools support exploration, reflection, and exchange.
4.1 Design Goals
- Wistful’s design goals are to support curiosity-driven exploration, capture what readers notice and pursue, and enable reflection on and exchange of traces.The goals draw on findings that exploratory reading can surface unanticipated questions and that readers reconstruct browsing through both moments and broader path structures [2].Source: §4.1 · p. 5
4.2 System Overview
- The system presents 771 UIST papers from 2020–2025 as a shared landscape and records ordered visits, transitions, and reading notes as traces.Readers explore on the Globe or Map, revisit and annotate in My Trace, and can publish traces through the Trace Library; traces remain private until published.Source: §4.2 · p. 5
4.3 Constructing the Scholarly Landscape
- The landscape uses title-and-abstract embeddings projected onto a sphere, with semantic similarity shaping proximity and corpus relationships supplying citation and authorship links.Seven k-means clusters provide topical landmarks, and the corpus is a bounded common space rather than a representation of HCI as a whole.Source: §4.3 · p. 5
4.4 Exploring and Reading the Landscape
- Readers navigate equivalent Globe and Map views, using spatial proximity and scholarly links as primary discovery routes alongside lightweight title or author lookup.Paper and author pages expose abstracts, papers, and relationships within the corpus, which readers can follow or use to return to the landscape.Source: §4.4 · p. 5
4.5 Capturing Creative Reading Traces
- Traces combine an ordered temporal record with a spatial path, while Field Notes attach readers’ reflections to individual visits.Revisits remain distinct; off-corpus pages can appear in the temporal record but not as spatial arcs.Source: §4.5 · pp. 5–6, Figure 2 · p. 6
4.6 Reflecting on Creative Reading Traces
- My Trace offers Path, Tree, and Graph views to revisit ordered visits, branching links, and relationships or interpretive connections.Readers can scrub through moments, distinguish contemporaneous Field Notes from retrospective Reflection Notes, and crop a contiguous segment for possible publication.Source: §4.6 · p. 6
4.7 Sharing Reading Traces
- Readers can publish a titled, framed segment as a fixed snapshot, then others can replay it, inspect notes, overlay it with their own path, and comment.Published traces can also accumulate as visible marks on the shared landscape.Source: §4.7 · pp. 6–7, Figure 3 · p. 7
4.8 Implementation Details
- Wistful precomputes its shared landscape and routes scholarly links and trace hops at runtime over a spherical Delaunay graph.A backend logs events and stores published traces as fixed snapshots that can be replayed independently of live browsing state.Source: §4.8 · p. 7
5 Method
The study used Wistful as a probe to examine how junior and senior HCI researchers explore literature, comparing it with their usual workflows. Researchers completed short open-ended tasks, and the team analyzed interaction traces alongside think-aloud and interview data.
5.1 Study Design
- A mixed-methods, within-subjects study compared 16 HCI researchers’ usual literature exploration with Wistful in two 15-minute tasks.Tasks began from assigned topics but left the path and resulting inquiry open; think-aloud and follow-up interviews probed participants’ judgments.Source: §5.1 · p. 7, §5 · p. 7
- Task framing varied by career stage: juniors developed a possible class-project inquiry, while seniors identified one to pursue with a student.Junior participants received more familiar topics and senior participants less familiar ones; career-stage differences are therefore descriptive, not causal.Source: §5.1 · p. 7
5.2 Recruitment & Participants
- The junior and senior cohorts each included eight researchers, classified primarily by HCI research experience.Participants were recruited through research-community channels, professional networks, and snowball sampling.Source: §5.2 · p. 8
5.3 Procedure
- Participants completed both tasks in counterbalanced order over approximately 90-minute Zoom sessions, sharing their screens and thinking aloud.After each task, they stated the inquiry that emerged and what led them to it.Source: §5.3 · p. 8
5.3.1 Wistful.
- Before exploring with Wistful, participants practiced with the system, then published their traces and reviewed and commented on others’ traces.The number and selection of prior traces varied with study order; early juniors were directed to senior traces.Source: §5.3.1 · p. 8
5.3.2 Usual Workflow.
- For the comparison task, participants used their usual discovery tools and strategies, including scholarly search engines, web search, and AI assistants.The usual-workflow condition served as an ecologically grounded reference rather than a standardized control.Source: §5.3.2 · p. 8, §5.1 · p. 8
5.3.3 Post-Task Questionnaire and Interview.
- Post-task questionnaires and a semi-structured interview compared exploration experiences and probed the judgments shaping them.Senior interviews also addressed mentorship and the development of scholarly sensibilities.Source: §5.3.3 · p. 8
5.4 Data Collection and Analysis
- Three authors open-coded transcripts alongside traces, notes, and logs, using logged behavior to establish actions and participant accounts to interpret them.Descriptive measures characterized search diversity, movement through the landscape, and note-taking; themes were refined through discussion across participants.Source: §5.4 · p. 8
6 Findings
Researchers approached the shared scholarly landscape differently, with familiarity, experience, and personal interests shaping what they pursued. Wistful’s traces helped readers orient during exploration and made paths available for reflection and comparison, though a trace alone did not always explain a reader’s choices.
6 Findings
- Reading paths reflected different personal reference frames, and traces made those differences visible while supporting orientation, reflection, and exchange.Post-task ratings were broadly comparable across workflows, with Wistful rated more favorably on enjoyment, reuse, record usefulness, and effort.Source: §6 · p. 8
6.1.1 Search Reflected Existing Research Practices.
- Among searchers, senior researchers had higher median query entropy than juniors (2.20 vs. 0.59 bits) and more often searched for authors, papers, or systems.Four of five senior searchers searched for an author by name, compared with none of the junior searchers; Wistful’s title-and-author-only search also constrained broader topic queries.Source: §6.1.1 · p. 9, Figure 5 · p. 10
6.1.2 Familiar Landmarks Provided Footholds.
- Familiar landmarks gave 11 of 16 participants footholds, while 12 of 16 questioned or probed the map’s organization against their own understanding of the field.Juniors more often began from papers they had read or authored; all eight seniors and four juniors questioned the map’s organization.Source: §6.1.2 · p. 9, §6.1.3 · p. 9
6.2 Familiarity and Personal Interests Influenced the Shape of Scholarly Dérive
- Familiarity and personal interests pulled readers beyond assigned topics, with senior departures often matching a familiar alternative and some junior departures reflecting interests not captured by familiarity ratings.Some seniors deliberately read along boundaries from familiar to less familiar areas; J5’s sewing interests and S7’s crochet interest also shaped paths across research boundaries.Source: §6.2 · p. 9, §6.2.1 · p. 9, §6.2.2 · pp. 9–10
6.3 Readers Composed Paths Across Semantic and Citation Structures
- Map transitions had a median geodesic distance of 0.30 radians, compared with 0.52 for hyperlinks, which also crossed continent boundaries more often (45% vs. 27%).The map organized papers by semantic proximity, while hyperlinks exposed citation relationships that could span distant regions; readers shifted between the two structures.Source: §6.3 · pp. 10–11, Figure 8 · p. 11, Figure 9 · p. 11
6.4.1 Traces Supported Orienteering During Exploration.
- Readers revisited My Trace to track where they had been and reorient, while Field Notes preserved more reader-authored context than baseline notes.61% of Field Note lines were in participants’ own words, compared with 25% of baseline document segments; 71% of baseline segments were labels, links, or references.Source: §6.4.1 · p. 11, §6.4.2 · pp. 11–12
6.4.3 Revisiting Traces Pointed Toward a Longer Time Horizon.
- Participants saw persistent Reflection Notes and accumulated traces as more useful over longer periods than during a single short session.They imagined returning after a break, reflecting on older reflections, visualizing research taste, and continuing research from a trace.Source: §6.4.3 · p. 12
6.4.4 Shared Traces Supported Comparison and Exchange.
- Shared traces made differences between readers’ paths visible and could point readers to unfamiliar papers, but traces alone did not always convey the author’s process.Participants differed on whether to share paths or paper lists, and the study-prompted example of a useful trace is not evidence of spontaneous exchange.Source: §6.4.4 · p. 12
7 Discussion
The discussion finds that researchers’ distinct scholarly sensibilities shape how they read, and that creative reading traces make some of those differences available for reflection and exchange. It also identifies limits: brief traces preserve movement more readily than the judgments behind it.
7 Discussion
- Wistful traces made visible plural reading orientations shaped by experience and interests, supporting orientation, recollection, comparison, and exchange.Participants mainly used their own and others’ traces for these purposes.Source: §7 · p. 12
7.1 From Sensemaking to Sensibility-Making
- Scholarly sensibilities encompass disciplinary expertise and wider experiences and interests that shape what researchers notice and pursue.Familiarity and everyday interests appeared in reading paths, while sensibility could guide reading before a researcher could articulate an interest.Source: §7.1 · pp. 12–13
7.2 Scholarly Dérive Under Pressure
- Even in 15-minute tasks, participants followed personal interests and familiar terrain into unplanned directions, but developing those directions can require more time.Participants connected long-term experience to making connections and worried that deadlines and AI-mediated acceleration could constrain exploration, especially for junior researchers.Source: §7.2 · p. 13
7.3 Making the Scholarly Map Come to Life
- The map served as both a computational representation and terrain for readers’ interpretations; 12/16 participants questioned or probed its organization.Citation links offered routes across semantic distance, while traces recorded readers’ paths over these overlapping structures (Fig. 10).Source: §7.3 · p. 13, Figure 10 · p. 14
7.4 Creative Reading Traces as Material for Reflection and Exchange
- Shared traces let researchers compare and learn from one another’s readings, but the rationale behind a path was harder for others to inherit.Participants imagined traces alongside conversation and explanation; they had not been asked to curate published traces for another reader.Source: §7.4 · p. 14
7.5 Limitations & Future Work
- The study captures brief expressions within a fixed UIST corpus and 15-minute tasks, so it cannot establish causal claims about career stage or sensibility development.The authors propose longitudinal study of how traces accumulate and support reflection over time.Source: §7.5 · p. 14
8 Conclusion
Wistful shows how scholarly sensibilities become visible in researchers’ reading paths and how creative reading traces preserve these expressions for reflection and exchange. The authors argue that scholarly systems should support readers in recognizing and developing their own ways of seeing.
8 Conclusion
- Researchers bring more to reading than an information need; their own ways of seeing shape the paths they make through literature.The conclusion frames those orientations as scholarly sensibilities.Source: §8 · p. 15
- Creative reading traces preserve fleeting expressions of scholarly sensibilities for reflection and exchange.Source: §8 · p. 15
- As scholarly systems model readers and personalize what they encounter, the authors argue for supporting readers in recognizing and developing their own ways of seeing.Source: §8 · p. 15
A.1 Formative Study Participants
Table 1 documents formative-study participants’ backgrounds, Wikipedia use, rabbitholing frequency, and starting articles.
A.1 Formative Study Participants
- Table 1 covers participants’ backgrounds.Source: Table 1 · p. 18
- The table includes participants’ Wikipedia use and rabbitholing frequency.Source: Table 1 · p. 18
- The table lists participants’ starting articles.Source: Table 1 · p. 18
A.2 Wistful Study Participants
Table 2 lists participant backgrounds and task assignments, including how junior and senior researchers were assigned continents by familiarity.
A.2 Wistful Study Participants
- Table 2 documents participant backgrounds and task assignments.Source: Table 2 · p. 18
- Participant IDs beginning with J denote junior researchers, and IDs beginning with S denote senior researchers.Source: Table 2 · p. 18
- Junior researchers received their two most familiar continents, while senior researchers received their two least familiar continents.Assignments were based on pre-study screener ratings.Source: Table 2 · p. 18
A.3 Post-Task Reflection Questionnaire
The post-task questionnaire was completed after each of two exploration tasks, with forms adapted for junior and senior researchers. Closed-response items generally used seven-point scales, alongside task reflections and open-ended questions.
A.3 Post-Task Reflection Questionnaire
- Unless otherwise noted, closed-response items used seven-point scales for familiarity and agreement.Familiarity ranged from 1 (Not at all familiar) to 7 (Extremely familiar); agreement ranged from 1 (Strongly disagree) to 7 (Strongly agree).Source: §A.3 · p. 18
A.3.1 Shared Header
- The shared header recorded participant ID and task order, then distinguished Usual Workflow from Wistful.Participants completed the corresponding task block after each task and paused after the first task before continuing.Source: §A.3.1 · p. 18, §A.3 · p. 18, §A.3.2 · p. 19
A.3.2 Junior Researcher Form
- The senior reflection form also asked whether participants knew the topic beforehand and afterward.Both forms included a question about the topic explored.Source: §A.3.2 · pp. 18–19
- Junior-form reflection items asked about ease, expressing interests, effort, effort’s worth, enjoyment, and willingness to reuse each workflow.Participants completed this block once for each task.Source: §A.3.2 · p. 19
- Junior-form open responses asked about usual exploration, workflow helpfulness or interference, Wistful’s real-world uses, and possible next steps.Participants were also invited to share additional comments.Source: §A.3.2 · p. 19
A.3.3 Senior Researcher Form
- The senior-researcher form adapted the topic, familiarity, and Likert structure to a mentoring scenario.Its reflection items covered inquiry depth, novelty, material usefulness, ease, effort, enjoyment, and reuse.Source: §A.3.3 · p. 19, §A.3.2 · p. 19
- Senior-form open responses asked how the workflow compared with advising a student and which Wistful features helped or interfered.Further prompts asked about real-world use, next steps, adapting produced materials, and other comments.Source: §A.3.2 · p. 19, §A.3.3 · p. 20
A.4 StitchFlow Example
The appendix shows one example of citation links crossing topic areas and maps spatial traces for all 16 participants by assigned continent.
A.4 StitchFlow Example
- Selecting StitchFlow exposed Delaunay-routed citation links from GenAI Creativity into Fabrication.Figure 11 shows what S7 saw after selecting StitchFlow on the map.Source: Figure 11 · p. 21
A.5 All Participants’ Spatial Traces
- Figure 12 groups all 16 participants’ spatial traces by their assigned continent.The listed groups span AR Interaction & Authoring, Fabrication, Sensemaking Tools, GenAI for Creativity & Design, Interaction in VR, and GenAI for Programming.Source: Figure 12 · p. 22
- The map’s assigned-continent groupings include six research areas, from AR Interaction & Authoring to GenAI for Programming.Figure 12 identifies the participant IDs within each group.Source: Figure 12 · p. 22